Portfolio Calculator: Metrics, Frontier, DCA & Retirement

The Portfolio Lab is a browser workbench for arithmetic on return or price series. Five panels — Metrics, Efficient Frontier, DCA Backtest, Retirement and Rebalancing — share one input format: one series per line, as periodic percentage returns or as a price path. Metrics computes annualised statistics, drawdowns, a correlation matrix and a beta/alpha regression; the frontier solves the mean-variance problem in closed form; the two backtests replay one history; retirement simulates a withdrawal plan. Everything runs locally; every output is an estimate from your inputs.

Updated 2026-10-01Reviewed by CalcX Editorial Team5 sources5 min read

How to run a panel

Open a panelfrom the tab bar: Metrics, Efficient Frontier, DCA Backtest, Retirement or Rebalancing.
Paste a series, one line per asset (Equity: 1.2, -0.4, 0.7), as returns in percent or a price path, and pick the frequency (daily 252 through annual 1). Metrics and Frontier also take a risk-free input; the demo button reloads the bundled inputs.
Set the assumptionstarget return and shorting switch (Frontier); contribution (DCA); weights, interval, drift band and one-way cost (Rebalancing); portfolio, withdrawal, horizon, return model, strategy, paths and seed (Retirement).
Read the grid, then the detail belowcorrelation matrix, equity and drawdown curves, frontier weights, percentile bands and ruin histogram.

Worked readouts

Every figure below comes from the bundled demo inputs.

Panel	Input	Readout
Metrics	24 monthly returns, risk-free 2%/yr	CAGR 11.77%, volatility 13.08%, Sharpe 0.766, Sortino 1.538, maximum drawdown −6.20%, Calmar 1.899, 95% VaR/CVaR 5.62% / 6.00%
Frontier	the same returns, annualised	minimum variance 5.86% at σ 0.93%, weights 13.7 / 87.5 / −1.3%; long-only 5.93% at σ 0.93%, weights 13.9 / 86.1 / 0.0; long-only max Sharpe 4.28 at 6.11%, σ 0.96%
DCA	prices 100 → 135, 1000 per period	invested 20,000, 203.259 units, final value 27,440 (+37.20%); lump sum 27,000 (+35.00%); advantage +440; average cost 98.397 vs a mean price of 101.05
Retirement	1,000,000; 40,000/yr; 30 years; 7% mean; 15% vol; 2.5% inflation; 2,000 paths; seed 20240101	success 73.7%; 4% comparison 73.7% (the first-year withdrawal is exactly 4.00%); median terminal 1,232,154; p10 0; p90 7,010,951; median withdrawals 1,756,108; median ruin year 23
Rebalancing	100,000; 60/30/10; every 3 periods; 5% band; 10 bp one-way	buy and hold 120,075; periodic 119,978 (8 rebalances, cost 9); the 5% band never fired, so the threshold rule matches buy and hold

The mathematics behind the panels

For weights w, expected returns μ and covariance Σ, the minimum-variance portfolio is w = Σ⁻¹1 / (1ᵀΣ⁻¹1) and the tangency portfolio is w ∝ Σ⁻¹(μ − rf1); the frontier follows from A = 1ᵀΣ⁻¹1, B = 1ᵀΣ⁻¹μ, C = μᵀΣ⁻¹μ and D = AC − B², with σ²(τ) = (Aτ² − 2Bτ + C) / D. Inputs are annualised by multiplying the mean and the covariance by the period count P, so the demo CAGR of 11.77% enters the frontier at 12.00% — arithmetic, not compounded.

Assumptions and limits

In-sample arithmetic.: Every statistic comes from the series you paste; change one input and all readouts change. The Monte Carlo replaces history with your model, drawing periods independently, so it understates clustered markets.
Annualisation follows the frequency selector.: Ratios scale by √P, expected returns by ×P and volatility by √P; a monthly series marked as daily misstates volatility by roughly a factor of 4.6.
Conditions and precision.: A singular covariance matrix has no unique frontier; no tangency portfolio exists when the risk-free input equals the minimum-variance return; the long-only solver is numerical and reports a residual; 2,000 paths give a success-rate standard error of at most about 1.1 points.
What it does not do.: It is an educational workbench: it fetches no market data, forecasts nothing, and recommends no rate, product or institution. This page provides general information and scenario estimates based only on the inputs you provide; it is not investment, tax, legal or other professional advice. Use Finance for TVM and cash flows, Statistics for raw samples, or the scientific calculator for one expression.

The Sharpe ratio is built per period

The Metrics panel reports a Sharpe ratio of 0.766 for the demo series, yet (CAGR − 2%) / 13.08% gives 0.747. The displayed ratio is assembled per period: the annual risk-free input is converted geometrically (2% becomes 0.165% per month), subtracted from every monthly return, and the mean of those excess returns is divided by their sample standard deviation before the ratio is scaled by √12. The naive version divides one annualised return by one annualised standard deviation, mixing a compounding view with an arithmetic one. Match the definition before comparing ratios across tools.

Where it fits

Reading a return series

Paste a fund, index or strategy series; Metrics returns the annualised block, equity and drawdown curves, a correlation matrix, and beta/alpha against any series in the same paste. Send an equity or drawdown vector to the data workspace.

Backtesting a contribution or rebalancing rule

DCA compares a fixed amount per period with a lump sum on one price path (average cost, units, IRR, final value); Rebalancing runs buy-and-hold, periodic and threshold versions of one weight target with turnover, cost and drift. Both are single-sample arithmetic; a rule that wins on one history need not win on another.

Stress-testing a withdrawal plan

Retirement reports a success rate, terminal percentiles, a ruin-year histogram and the 4% comparison. On the demo inputs the guardrail strategy moves success from 73.7% to 99.8% and median withdrawals from 1,756,108 to 1,594,979, because spending adjusts along the way.

References

Wikipedia, Modern portfolio theory, en.wikipedia.org （访问日期：2026-10-01）— mean-variance optimisation and closed-form frontier solutions.
William F. Sharpe, The Sharpe Ratio, The Journal of Portfolio Management, 1994, web.stanford.edu （访问日期：2026-10-01）— the ex-post ratio and its time dependence.
Wikipedia, Dollar cost averaging, en.wikipedia.org （访问日期：2026-10-01）— average cost and the harmonic mean identity.
Wikipedia, Retirement spend-down, en.wikipedia.org （访问日期：2026-10-01）— withdrawal rates and the 4% rule of thumb.
Wikipedia, Monte Carlo method, en.wikipedia.org （访问日期：2026-10-01）— simulation and sampling error.